A NNALES DE L ’I. H. P., SECTION A
V. I RANZO J. L LOSA F. M ARQUÉS A. M OLINA
An exactly solvable model in predictive relativistic mechanics. I
Annales de l’I. H. P., section A, tome 35, n
o1 (1981), p. 1-15
<http://www.numdam.org/item?id=AIHPA_1981__35_1_1_0>
© Gauthier-Villars, 1981, tous droits réservés.
L’accès aux archives de la revue « Annales de l’I. H. P., section A » implique l’accord avec les conditions générales d’utilisation (http://www.numdam.
org/conditions). Toute utilisation commerciale ou impression systématique est constitutive d’une infraction pénale. Toute copie ou impression de ce fichier doit contenir la présente mention de copyright.
Article numérisé dans le cadre du programme Numérisation de documents anciens mathématiques
http://www.numdam.org/
1
An exactly solvable model
in Predictive Relativistic Mechanics. I
V. IRANZO, J. LLOSA, F.
MARQUÉS
and A. MOLINADepartament de Fisica Teorica, Universitat de Barcelona (*)
Vol. XXXV, n° 1, 1981, Physique ’ théorique. ’
ABSTRACT. 2014 We present a
family
of hamiltonian models forN-particle
systems in Predictive Relativistic Mechanics. The relation between
posi-
tion and
velocity
variables and thenon-physical
canonical ones is derived.The case of a harmonic oscillator
potential
is studied in detail. We compareour model with some
singular lagrangian
systemsalready appeared
inthe literature.
1. INTRODUCTION
This is the first of a series of papers about a class of systems of
particles interacting
at a distance and theirquantization.
The interest of relativistic action-at-a-distance theories hasrecently
grown[7] ] [2] ] [3] ] [4] ] [J] ] [6].
From the theoretical
viewpoint
this is due to the fact that the simultaneous character of the interaction has been madecompatible
with Poincare invarianceby
Predictive Relativistic Mechanics.Also,
hamiltonian sys-tems with constraints have been better understood. These systems are
only
defined on asubmanifold
of thephase
space and this enables one to circumvent the no-interaction theorem. The connexion between both(*) Postal address: Diagonal, 647, Barcelona-28, Spain.
Annales de l’Institut Henri Poincaré-Section A-Vol. XXXV, 0020-2339/1981/1/$ 5,00/
(Ç) Gauthier-Villars
2 V. IRANZO, J. LLOSA, F. MARQUES AND A. MOLINA
formalisms,
Predictive Relativistic Mechanics andSingular Lagrangian Theory
has beenrecently
establishedby
the authors[7] ] [8 ].
From a
phenomenological viewpoint
the use of covariant harmonicoscillator models has allowed to obtain the mass spectrum of hadrons and has
provided
a theoretical way toexplain
thequark-confine-
ment
[9] ] [10 ].
Before the work of Droz-Vincent
[5] ] only perturbative
solutions ofphysically interesting problems
were known[7] ] [79] ] [20] ]
within the framework of Predictive Relativistic Mechanics.We present here a
family
ofexactly
solvable models which includesas a
particular
case for twoparticles
the Droz-Vincent system.Although
we do not believe that this kind ofsimple
models can describethe interaction between
quarks,
this will beonly possible
in the frameworkof a
relativistic
secondquantization. Nevertheless,
theclearness,
the sim-plicity
and the exactsolvability
of these modelspermit
a better under-standing
of the main features ofquark
interaction.In sec. 2 we present a short reviews of Predictive Relativistic Mecha- nics and the
general
characteristics of our model. It isinspired
on Dirac’sidea
[77] ]
ofdistinguishing
the kinematic generators from thedynamic
ones among those of the Poincare
Group.
In our case we make the above mentioned distinction among the generators of theComplete Symmetry Group
which is an abelian extension of the PoincareGroup.
In sec. 3 we present the model in detail. In order to circumvent the No- Interaction Theorems
[72] ] [7~] ]
we deal with anon-physical
canonicalcoordinates. The relation between the
position
andvelocity
variablesand the canonical ones is derived in sec. 4. In sec. 5 we present some
appli-
cations of the model: harmonic oscillator
potential,
freeparticles
system andcomparison
with somesingular lagrangian
models.2. PREDICTIVE HAMILTONIAN SYSTEMS
A Predictive Poincare invariant system is a second order differential system
which describes the
dynamics
of Nparticles.
The acceleration
0~
mustsatisfy [7~] ] [7~] :
Annales de l’Institut henri Poincaré-Section A
where the summation convention holds for c, ... ...
indices,
and the metric is taken(2014+++).
Equation (2.3)
exhibits that 6a is the proper time of theparticle
« a »1 appart from a
multiplicative
constant which we shall takeequal
to 2014.If ~a were the proper
time,
it wouldyield
~and the coordinates
(x, 7~)
would nolonger
beindependent.
This wouldobstruct the construction of a halmitonian formalism. On the contrary, the choice of mentioned above allows
~a
to take on any valuewhich, according
to eq.(2.2),
will be anintegral
of motion.Equations (2.3)
show that the solution xa of system(2.1) only depends
on the parameter 6a.
Furthermore, they
guarantee theintegrability
ofthe system.
Equations (2.3)
can also be written as :The invariance of
(2 .1 )
under the Poincare groupimplies
that the func-tions
8a(x, ~c)
behave as four-vectors invariant under translations. This is e uivalent towhere :
are the infinitesimal generators of the Poincare
Group.
Equations (2.4)
and(2.5)
exhibit thatHa, PJl’
span a realization of an abelian extension of the Poincarealgebra [16 ].
The associated trans- formationgroup ~
is the directproduct
of the Poincare group(purely kinematic)
and thedynamical
group [RN.A hamiltonian formalism for the system
(2 .1 )
is determinedby
a sym-plectic
form Q on invariant under J 8>AN.
As it is well
known,
a functionf
can be associated to each field A that leaves Q invariant. This function is definedby [16 ].
This function is defined at least
locally
on and it isunique
except for anarbitrary
additive constant.4 V. IRANZO, J. LLOSA, F. MARQUES AND A. MOLINA
In
particular,
the functions associated to andHa
are the linear momentum, theangular
momentum and thehamiltonians, respectively.
This is the way to follow in order to
quantize
a classical system like(2.1) [17].
We mustpoint
out that thisprocedure
involves pertur- bative solutions as in the cases ofelectromagnetic
interaction and of short range scalar and vector interactions[18 ].
We will
follow, however,
Droz-Vincentsapproach [5 ].
Instead ofconstructing
a hamiltonian formalism for agiven
interaction(e.
g. elec-tromagnetism...),
wedisplay
someexactly
solvable models which serve us to know better the structure of these theories. We also want some of these models to describe the interaction betweenquarks.
We work in an
adapted
canonical coordinates systemp~)
ofQ[16 ],
i. e. : these coordinates
satisfy
thefollowing
conditions :i ) q~
andpb (a,
b =1,
...,N)
are four-vectorsand qa
behaves like~
under translations.
ii)
thesympletic
form Q can beexpressed
as :or,
equivalently,
the Poisson bracket associated to Q is such that :Then, if
Pu
andJ 111-
are the functions associatedby
eq.(2.7)
to the gene- ratorsPl1
andof !J>,
weimmediately
obtain from the condition(i)
that :Therefore,
thedynamical
aspects of the system will be present in the functionsHa(q, p) (associated
toHa)
and in the functionsx~(q, p), ~(~, p)
which relate the
position
andvelocity
variables to the canonical coor-dinates and momenta.
Hence,
a model will consist ingiving
N functionp) satisfying
thefollowing
conditions :i)
eachHa( q, p)
is invariant under the Poincare groupBoth conditions guarantee that eq.
(2.4)
and(2.5)
will be fulfilledby
the
generators Ha
associated toHa(q, p) by (2.7).
After that several
problems
still remain. First we have to derive therelationship
between theadapted
canonical coordinates(q, p)2014which
have not any
physical significance and
theposition
andvelocity
variables
~).
As far as we do not know these relations our hamil- tonian model will neither be able topredict anything
nor to becompared
with other
physical
models.l’Institut Henri Poincaré-Section A
Takin
into account eq .( 2 .1 ),
the functions x 03B1 mustsatisfy :
This means that the evolution of
xa depends only
on the parameter 6a, unlike the canonicalcoordinates q a
whichdepend
on the whole set of parameters ... ,6N~.
.
If we can solve eq.
(2.11)
we will know the relationp).
As we shallsee later a
good
set of initial data willpermit
us to collect one among theinfinity
of solutions of(2.11).
Secondly, according
to(2.2) {H~ ought
to be an inte-gral
of motion. It isextremely complicated
toimpose
this condition.However we can circumvent this
difficulty by assuming
that the fieldassociated
trough
Q top)
is notequal
toHa
butonly proportional.
That is to say, if we write :
the parameter 03C4a is no
longer equal
to the proper timeNow, given
a solutionxa
=p) (a
=1,
...,N)
of eq.(2.10),
we needonly
torequire
thechange
of variablesto be invertible.
There is at last one
problem
left : what is the relation between(x, x)
and
(x, 7r)? As x a and 03C0 a
areparallel
weonly
need the ratio between theirlengths
in order to defineprecisely
thechange
of variables. This can be doneby giving
a fixed value to eachintegral
of motion ...,HN . Similarly
to the case of freeparticles
or ofseparable
interaction[7~] ]
we shall take
The relation between the parameters 6Q
and 03C4a
can beeasily
obtained :where C. I. means the set of initial conditions which determine the
trajec-
tories. The
change
of parameter can be now obtained from(2.15) by
aquadrature.
3. A FAMILY OF MODELS
We shall choose the hamiltonians
Ha
with thefollowing
form :where :
Ta
is aquadratic
functionof pb,
V is a Poincare invariant function and it is the same for all a.6 V. IRANZO. J. LLOSA, F. MARQUES AND A. MOLINA
For the sake of
simplicity
we shall alsorequire
that bothTa
and V are well behaved underparticle interchange.
That is to say, if 6 ESN (the symmetric
group of Nelements),
then :This condition
implies
that :where
03B11, 03B21,
7iand 03B41
1 arearbitrary
constants.Instead of
using
the variablesp~)
it is more suitable toperform
thecanonical transformation :
c ~ r, , ... , i .
The new variables P, y03BDAm z03BBB are translation invariant.
Hence, V(q, p)
mustbe some function of the
following
scalarvaria-bles:
where
ZB
are theprojections of y A, Za perpendicular
to Pu :In terms of the new
variables,
eq.(3.2) yields :
where,
in orden tosimplify
thenotation,
we have introduced :and (x,
~3, y, ~
arearbitrary
constants.The functions
Ha
that we have chosen mustsatisfy
thepredictivity equations (2.10).
As the hamiltoniansHa
are well behaved underparticle interchange
it suffices torequire : {
= 0 A =2, 3,
..., N.Using (3 . 6)
these conditions
yield :
Annales de I Henri Poincaré-Section A
Taking into account that V depends on the variables
(3 . 4)
we obtain :The
simplest
models are those with b = 0. Otherwise the differential system(3. 7)
becomesextremely complicated.
For thesesimplified
modelseq.
(3 . 7) yields :
Hence,
if we choose 5 = 0 andV(P2, PyA, AB, AB, AB)
the pre-dictivity
condition(2.9)
isautomatically
fulfilled.Since V does not
depen
on the functionsPyA
areintegrals
ofmotion.
And,
as p2 is also anintegral
of motion theproducts (PpJ
areconserved
quantities,
too.Other
integrals
of motion are those associated to the generators of the Lorentz group :which can be
expressed
as :with :
where : M =
( - P2) 1 ~2
and is the four-dimensional Levi-Civita tensor.Analogously
to the canonical transformation(3 . 3)
weperform
achange
of parameters :
It
immediately
follows from eq.(3.12&)
thatYÀ
andzA
do notdepend
on
ÀB
andthey
aresolely
functions of ~,.Therefore,
the transverse internal motion of the system(i.
e.: the evolution of the relative coordinates andVol. 1-1981.
8 V. IRANZO, J. LLOSA, F. MARQUES AND A. MOLINA
momenta in the
orthogonal hiperplane
to isgoverned by
theordinary
second order differential system :
The
potential
V candepend
on~ 2~
but also on p2 andHowever,
as p2 and
PyA
areintegrals
of motionthey
can beregarded
as parameters in order tointegrate
the system(3.13). Therefore,
eqs.(3.13)
allows us toobtain
yA(~,;
C.I.)
and2À(À;
C.I.) independently
of the evolution of the other variables.Finally,
the evolutionequations
of(PX)
and(PzA) yield :
~ A
If V does not
depends
on p2 andPyA,
the system(3 .13)
would be imme-diately integrable,
because p2 andPyA
are conservedquantities.
As weshall see
later,
if we know(PX)
and(PzA)
in termsof ..1, ~,A
and the initialconditions then we can
easily integrate
theposition equations (2 .11 ).
Therefore we shall consider hereafter
only potentials
of the form :zc).
In this case,
by integration
of eq.(3 .14)
we obtain :The hamiltonians to be considered hereafter are therefore:
4. INTEGRATION OF THE POSITION
EQUATIONS
In order to
complete
the modelproposed
in the last section we mustintegrate
theposition equation (2.11).
By analogy
with thetwo-particle
model of Droz-Vincent[5] and
withAnnales de l’Institut Henri Poincaré-Section A
some
singular lagrangians [2] ] [3] ] [4] we
shall take thefollowing
initialconditions :
The Frobenius theorem
guarantees
us that there exists aunique
solution of
(2 .11 ) satisfying
the initial conditions(4 .1 ).
Given onepoint
there exists a
unique integral cpa(~1, ..., ’~N),
...,
TN)
of thegeneralized
Hamiltonequations depending
on N para- meters :Satisfying ~(0.. ~ 0) = ~
...,0)
=p~.
By
variation of all theparameters
but one za, thisintegral
submanifoldpermits
us to go fromQ
toQa
=p~)
where :
bza
= 0 and are suitable variation of theparameters.
From eq.
(2.11) p)
is not alteredby
achange
of03C4a,(a’ ~ a).
Hencexa(Q)
isequal
to andusing
eq.(4.1)
we have:in terms of
Q
=(q~(Ü)’
Therefore,
if we can express in terms ofQ
=(~S, p~)
the inte-gration
of(2.11)
with the initial conditions(4.1)
will be concluded.From eq.
(3.3), (3.11)
and(4.3)
we obtainwhere the subindex has been omitted from
Ku, M,
pv and becausethey
areintegrals
of motion.Writing
eq.(3.15)
atQ
andQm substracting
them andtaking
into accounteq.
(4.1),
we have :where
~~,A, ~~,
areexpressed
in terms of the(N - 1)
parametersby (3.12).
Vol. XXXV, n° 1-1981.
10 V. IRANZO, J. LLOSA, F. MARQUES AND A. MOLINA
We can find the unknowns
(PX)
andU
from eq.(4 . 5)
and we obtain :where the subindex
Q
is understood in thequantities
on theright
hand side.Now the
quantities (~ 2014
are still to be found. Let us write(3.13)
as :I.
C.), ~;
I.C.)}
be thegeneral integral
of(4 . 8). Then,
it is
quite
clear that :where 5x is
given by
eq.(4.7).
Substituting
eq.(4.9)
and(4.6)
in eq.(4.4)
theintegration
of theposi-
tion
equations (2.11)
iscomplete.
We have topoint
out thatonly
theterm
(~ 2014 depends
on thepotential.
SinceQ
isarbitrary,
the resultis
general,
so the indicesQ
and(0)
will bedropped
from now on.The
change
of variables( q, p)
--+(x, x)
can befinally
written as :were the subindex
(0)
isdropped
from(4.9).
If this
change
of variables is to be invertibleand acb
and pl1 wellbehaved,
the followins conditions must be also satisfied :
In any case ’ we can choose o
~,
y and 0 V such that the conditions above hold 0 in(TM4)N
except,perhaps,
in a set of measure ’ zero.Annales de l’Institut Henri Poincare-Section A
5. APPLICATIONS a~ Harmonic oscillator.
In order to
explain
the interaction betweenquarks,
differentmodels of relativistic harmonic oscillators have
appeared
in the litera- ture[27] ] [22] ] [5] ] [6 ].
We shall now present a modelof N-particle
har-monic
oscillator,
which for N = 2 includes the model of Droz-Vincent[5] ]
mentioned above as a
particular
case. We shall use thepotential:
Then,
theequations
of motion for the relative transverse coordinates and momenta can be written as :Then,
we obtain thefollowing eq uation
In other
words,
all the relative transverse coordinates oscillate with the samefrequency
úJ =NN03B3K.
On the otherhand,
eq.(3.15)
showsthat the component
parallel
to pl1 of these coordinatesdepends linearly
on za.
To
completely
solve this model we must derive thechange
ofvariables
( q, p)
-~(x, x).
From(4 . 9)
and(5 . 3)
we obtaineasily :
where ~~. is
given by (4. 6).
The introduction of the results(5 . 4)
in eq.(4 .10)
a
and
(4. .11 )
determines the above mentionedchange
of variables. For theparticular
case N =2
a =1 - 1
-1
our hamiltoniansyield:
2 8 2Vol. XXXV, n° 1-1981.
12 V. IRANZO, J. LLOSA, F. MARQUES AND A. MOLINA
which coincides with the harmonic oscillator hamiltonian
presented by
1 Droz-Vincent
[5 ],
if we take :KDv
=2 K .
In this case the
position
andvelocity
variables can beexpressed
in aparticularly simple
form :where: y = y2 , z = z2 , r~a =
(- 1)~+ y a
=1,
2.Equations (5.6)
coincide with Droz-Vincent’sexpressions.
We haveto
point
outthat,
when restrictedto ~ _ ~ (TM4)2 ~
P -( q
1-
q2)
=0},
not
only x~( q, p)
= but also :p)
=p~.
On the other
hand,
if N > 2 it is nolonger possible
to find suitable values of (X,~/~~y
such thatT - - a 1 2 pa 2 . Because of this it was necessary to gene-
ralize the
expression
ofT~
in order to describe(in
a solvableway)
a modelinvolving
three or moreparticles.
b~ Singular lagrangian systems.
Most of the
singular lagrangian
systemsexisting
in the literaturepresent
the constraints :
We are
going
to showthat,
when N = 3 and for suitable values of (x,{3,
y andV,
our hamiltonian model is apredictive
extension of theTakabayasi lagrangian
system[4 ].
In otherwords,
we aregoing
to provethat,
when restricted to the submanifoldE*,
theequations
of motion derived fromour model
yield
theequations
of motion ofTakabayasi (*).
Each
integral
of ourgeneralized
Hamiltonequations (4.2)
is a N-sub-manifold
parametrized by
Ti, ...,TN which intersects E* in a curve.This is the world line of the system on L*. It is immediate to show that the derivative
a/~~,
is tangent toE*,
while(A
=2,
...,N)
are not.Therefore, ~?
is agood
parameter to describe the evolution of the sys-tem on E*. Hereafter we shall write instead of :E* and we shall consider a
potential
of the formV(zA)-we
have to realize that zA =zA
on E*.
(*) The same can be done easily for the Kalb and Van Alstine Lagrangian [2]
which has the same structure as (5. 10] with N = 2.
Annales de l’Institut Henri Poincaré-Sectíon A
The
equations
of motion(4.2)
restricted to E*yield :
from which we obtain the
following
second order differentialequations :
Let us now consider the
Takabayasi Lagrangian [4 ] :
where
(’)
means the derivative referred to any Poincare invariant para- meter T and ~’11 is theprojection
ofperpendicular
to the relative coor-dinates z2 and z3.
In
fact,
the modelpresented
in ref.[4] ]
does not exhibit the crossterm
(z2, z3).
We have added it in order to maintain the invariance underparticle interchange.
It is immediate to prove that there is a linear trans- formation whichchanges
theTakabayasi lagrangian
into the form(5.10).
From the
lagrangian (5.10)
we derive theequations
of motion :which hold on the constrained submanifold
= 0, PyA = 0, A = 2,3}.
Here pl1
and yA
are theconjugated
momenta of Xl1 andzA:
From eq.
[3] ] [8] ] [P] ] [77] ] [72] we
seeeasily
that both models coincide if we take :Vol. XXXV, n° 1-1981.
14 V. IRANZO, J. LLOSA, F. MARQUES AND A. MOLINA and
where and C are numerical constants and it is clear that is an
integral
of motion. We have topoint
out that eq.(5.14)
is theexpression
of
U(zA) given by Takabayasi
in order to obtain a freeparticle
system when V = 0.c)
Freeparticle system.
When V = 0 the
equations
of motion(4 . 2)
read :Hence,
thecoordinates q a depend linearly
on 03C4b and the momentap 03B1
areintegrals
of motion.We have to
point
out that the relation betweenp~ and qa
is not the usualsimple
one :qQ = p~ .
This is due to the fact that we have had to
give
up the usual freeparticle
form
of T - - 1 2 a
=1,
..., N.Nevertheless,
in terms of theposition
andvelocity
variables wecan see
immediately
from eq.(2.10) (4.10) (4.11)
and(5.15)
that :where oa and oa
are thepositions
and velocities at T 1 = ... = TN = 0.6. CONCLUSION AND OUTLOOK
In earlier models with interaction-at-a-distance
[2] ] [3] ] [4] ] [5] two
orthree
particle
systems have been treatedseparately. They
neither can begeneralized
toN-particle
systems norpermit
to deal withquark
interaction for mesons andbaryons
in a unified way. Here we havepresented
afamily
of
N-particle
systems which includes the earlierquoted
ones when N = 2or N = 3 it is a
predictive
extension of some of them[2] ] [3] ] [4] and
coincides with the
predictive
Droz-Vincent’s model[5 ].
Many
of those models areincomplete
because the relation between theposition
andvelocity
variables and the canonical ones is notspecified.
Hence, the
physical meaning
of the coordinates involved remains unknown.l’Institut l’Institut Poincaré-Section A
We have
carefully
discussed thisquestion
and the above mentioned rela- tion has beenexactly
derived for ourfamily
of systems.Owing
to thegenerality
of thepotential V,
our modelpermits
to describedifferent kinds of interaction. We have
applied
it to theparticular
case ofa harmonic oscillator
potential
since it isspecially interesting
in orderto describe the
quark
interaction.In a future work we shall
quantize
the model and we shall compare the mass-spectrum and otherproperties
with thephenomenological
dataobtained from hadrons.
[1] L. BEL, A. SALAS and J. M. SÁNCHEZ, Physical Review, t. D7, 1973, p. 1099.
[2] M. KALB and P. VAN ALSTINE, Yale Report COO-3075-146 (june 1976).
[3] D.
DOMINICI, J. GOMIS and G. LONGHI, Il Nuovo Cimento, t. 48B, 1978, p. 152.[4] T. TAKABAYASI, Prog. of Theor. Phys., t. 58, 1977, p. 1299.
[5] Ph. DROZ-VINCENT, Ann. Inst. H. Poincaré, t. 28A, 1977, p. 407.
[6] H. LEUTWYLER and J. STERN, Ann. of Physics, t. 112, 1978, p. 94.
[7] J. A. LLOSA, F. MARQUÉS and A. MOLINA, Ann. Inst. H. Poincaré, t. 32A, 1980,
p. 303.
[8] F. MARQUÉS, Doctoral Thesis. Universitat de Barcelona (september 1980).
[9] H. LEUTWYLER in « Quantum Cromodynamics », Lecture Notes in Physics,
t. 118, Springer-Verlag, 1980, p. 121.
[10] J. GASSER, H. LEUTWYLER, J. JERSAK and J. STERN, Berna Preprint, 1979.
[11] P. A. M. DIRAC, Rev. of Modern Physics, t. 21, 1949, p. 392.
[12] D. G. CURIE, T. F. JORDAN and E. C. G., Sudarshan, Review of Modern Physics,
t. 35, 1963, p. 350.
[13] H. LEUTWYLER, Il Nuovo Cimento, t. 37, 1965, p. 556.
[14] Ph. DROZ-VINCENT, Physica Scripta, t. 2, 1970, p. 129.
[15] L. BEL, Ann. Inst. H. Poincaré, t. 14, 1971, p. 189.
[16] L. BEL and J. MARTIN, Ann. Inst. H. Poincaré, t. 22, 1975, p. 173.
[17] L. BEL in Differential Geometry and Relativity. Éd. Cohen and Flato. Dor-
drecht, 1976, p. 197.
[18] C. BONA, Doctoral Thesis. Universitat Autonoma Barcelona (june 1980).
[19] R. LAPIEDRA, A. MOLINA, J. of Math. Phys., t. 20, 1979, p. 1309.
[20] R. LAPIEDRA, F. MARQUÉS and A. MOLINA, J. Math. Phys., t. 20, 1979, p. 1316.
[21] R. P. FEYNMAN, M. KISLINGER and F. RAVNDAL, Physical Review, t. D3, 1971, p. 2706.
[22] Y. S. KIM and M. E. Noz, Physical Review, t. D8, 1973, p. 3521.
(Manuscril reçu le 17 fevrier 1981)